Examples of large language models

• Google Bert, 2017
• ChatGPT, 2022
• Llama-2, 2023

Some analogies of LLMs

• Emily Bender et al.: Stochastic parrots
• Ted Chiang: “ChatGPT is a Blurry JPEG of the Web”
• Simon Willison: “Calculators for words”

A technical definition for LLMs

• LLMs model text and can be used to generate it
• LLMs receive large-scale pretraining
• LLMs make inferences based on transfer learning

How LLMs can help social scientists

• Writing assistants
• Preprocessing datasets, eg. by summarising or generating keywords
• Creation of synthetic datasets for piloting research or preserving privacy
• Categorising text or extracting particular features from it

General requirements for concepts in content analysis

Validating analysis requires people

• Core challenge for analysis is finding analytical concepts that
  1. Provide analytical value, i.e. reveal something insightful about data
  2. Are clear enough so that multiple people will agree on them
  3. Can be operationalised so that models such as LLMs can apply them
• Agreement between people is proxy for validity of concepts - for some research, codebook verified before modelling
• Finding right concepts may require many attempts, and LLMs may make iterative process easier

Typical workflow for modelling work

Typical workflow:

• You train your own model
• ..with an architecture specific to the problem
• ..and exclusively on your own data

Potential changes in workflows with LLMs

• LLMs are “Swiss army knives”: Single models can be applied for many different tasks
• A potential simplification in expertise required for modeling - but at a cost!

3 ways to do content analysis

Hate speech research for Cabinet Office

Definition of “hateful speech”

hateful speech: Messages that

1. Contains slur from pre-defined list
2. Use slurs used in abusive or threatening way

Creating a dataset for training

Finetuning an LLM for text classification

Finetuning is a process where an pre-trained model goes through a smaller training process to perform a specific task.

Finetuning a model in Python

model = AutoModelForSequenceClassification.from_pretrained(
    "TurkuNLP/bert-base-finnish-uncased-v1"
)

training_args = TrainingArguments(
    "nethate", evaluation_strategy="epoch", logging_steps=30
)
metric = load_metric("accuracy")

trainer = Trainer(
    args=training_args,
    model=model,
    train_dataset=train_dataset,
    eval_dataset=eval_dataset,
    compute_metrics=compute_metrics,
)
trainer.train()

finetuned_finbert = pipeline(
    model=model, tokenizer=tokenizer, task="sentiment-analysis", return_all_scores=True
)

Goal: A “boring”, boilerplate model

Categorisation results

Evaluating the performance of the classification

Examining how the model works

An alternative workflow: zero-shot annotation

• Can be done with LLMs that are trained to be directed with prompts
• Zero-shot: When the model is asked to perform a task that it hasn’t specifically been trained on, i.e. to apply new categories that are described to it
• Few-shot: When the model is provided with a limited number of examples for its task
• Results tend to be worse than with finetuned models

Example of zero-shot annotation

An example of annotating text: Assessing the amount of time passing in literary passages.

Prompting GPT-4 to read fiction

Read the following passage of fiction. Then do five things.
 
1: Briefly summarize the passage.

2: **Reason step by step to decide how much time is described in the passage. If the
passage doesn't include any explicit reference to time, you can guess how much 
time the events described would have taken**. Even description can imply the 
passage of time by describing the earlier history of people or buildings. But 
characters' references to the past or future in spoken dialogue should not count 
as time that passed in the scene. Report the time using units of years, weeks, 
days, hours, or minutes. Do not say zero or N/A.

3: If you described a range of possible times in step 2 take the midpoint of the
range. Then multiply to convert the units into minutes.

4: Report only the number of minutes elapsed, which should match the number in 
step 3. Do not reply N/A.

5: Given the amount of speculation required in step 2, describe your certainty 
about the estimate--either high, moderate, or low.

The passage follows: <omitting this to save space>

Results

1: A detective wakes up 'the next morning,' looks out a window for an undefined 
time, drinks (and presumably needs to make) two cups of coffee, then shaves and 
showers and gets dressed before stepping out his front door and seeing a car.

2: **Making coffee, showering, and getting dressed take at least an hour. There's 
some ambiguity about whether to count the implicit reference to yesterday 
(since this is 'the next morning') as time elapsed in the passage, but let's 
say no, since yesterday is not actually described. So, an hour to 90 minutes.**

3: 1.25 hours have elapsed. Multiplying by 60 minutes an hour that's 75 minutes.

4: 75 minutes.

5: Low confidence, because of ambiguity about a reference to the previous day.

Findings

“Watching the model think out loud about each passage actually gave me a better sense of how results were produced than I could extract from my own records of human research in 2017.”

General challenges of working with LLMs

• Most recent powerful models closed, proprietary, opaque
• Proprietary models make research hard to reproduce
• Industry influence, centralisation (GPT-4 cost 100 million to train)
• Difficult to interpret (though models can be prompted to explain themselves)
• Sensitive data cannot be analysed with cloud services
• Ethics of models, eg. sourcing of their training data

LLMs might help develop analytical concepts

Let’s try to access CSC notebooks

• Go to the URL https://notebooks.rahtiapp.fi
• Log in with Haka account
• Press “Join workspace” in top right
• Enter code jyu-07ippvlk
• You may press power button icon to start your own virtual machine
• After waiting a moment, you can press in the same location to “enter session”

Using Jupyter Labs

• After you have entered your session, you are inside of a system called Jupyter Labs, where you can view and execute Jupyter Notebooks
• On the left, you have list of files and folders. Double click on “jyu-llm-methods”
• There are two notebook files with filenames “Exercise_…”. Double click both to open them
• Note that the virtual machine only lasts for 12 hours, and not all folders persistent

2 exercises

1. Text classification with supervised learning and finetuning an LLM
   • Example data: Yelp reviews with positive/negative classification
   • Some techniques for evaluating results, such as confusion matrices and “scattertext” for false classifications
   • Tasks: E.g. try out with other data, examine biases
2. Zero-shot text annotation by prompting an LLM
   • Example data: Russian and British news articles about Ukraine
   • Sample prompts for extracting sources of information
   • Tasks: E.g. Try out other prompts, examine model’s mistakes
3. Text classification/annotation in ChatGPT
   • If you would rather not use the notebook interface or Python, feel free to think of applying other interfaces such as ChatGPT for your work

:: {.notes} - Spirit of exercises: above all, try out something that is useful for you - Here are some resources that you can work with :::